paper-with-me

Papers

Adaptive Head Budgeting for Efficient Multi-Head Attention

2026-04-24 · Bilal Faye, Abdoulaye Mbaye, Hanane Azzag, Mustapha Lebbah arxiv

Multi-head attention enables Transformers to capture diverse representations, but all attention heads are typically activated for every input, regardless of task complexity. For coarse-grained tasks such as text classification, where relevant information is often global, this fixed allocation can introduce unnecessary computation. We propose BudgetFormer, a Transformer architecture that dynamically allocates attention heads on a per-input basis. The model learns both a head budget and a relevance distribution to select the most informative heads. To support effective head selection, we introduce a training strategy that balances exploration and exploitation. Experiments on text classification tasks show that BudgetFormer reduces FLOPs and memory usage while matching or surpassing the performance of standard multi-head attention. These results highlight adaptive head allocation as an effective approach to improving Transformer efficiency and performance.

📄 PDF Abstract BibTeX arXiv:2604.22583

Code (0)

등록된 구현이 없습니다.

Tasks

Text Classification

Similar Papers 제목 키워드 기반

HARD-KV: Head-Adaptive Regularization for Decoding-time KV Compression

2026-06-27 · Yuxuan Yang, Feiyang Ren, Bowen Zeng, Dalin Zhang 외 arxiv

Long-context LLM inference faces a fundamental conflict: head-adaptive compression algorithms (e.g., Top-$p$ nucleus sampling) offer superior accuracy by dynamically fluctuating memory budgets, yet modern inference engin…

AdaTok: Self-Budgeting Image Tokenization with Quality-Preserving Dynamic Tokens

2026-06-05 · Xiaocheng Lu, Yuxi Chen, Jie Zhang, Jian Liu 외 arxiv

Image tokenizers, from 2D grids to recent 1D sequences, typically encode every image with the same fixed number of tokens. Yet visual complexity is highly heterogeneous, so a uniform budget overspends on simple inputs an…

Representation LearningImage Generation

Mask Tokens as Prophet: Fine-Grained Cache Eviction for Efficient dLLM Inference

2025-10-10 · Jianuo Huang, Yaojie Zhang, Yicun Yang, Benhao Huang 외 arxiv

Diffusion large language models (dLLMs) present a promising alternative to dominant autoregressive models (ARMs) by the ability of parallel decoding at the expense of substantial computation and memory costs. Specificall…

Adaptively Sparse Transformers

2019-08-30 · IJCNLP 2019 11 · Gonçalo M. Correia, Vlad Niculae, André F. T. Martins

Attention mechanisms have become ubiquitous in NLP. Recent architectures, notably the Transformer, learn powerful context-aware word representations through layered, multi-headed attention. The multiple heads learn diver…

DiversityMachine TranslationTranslation

An Efficient Hybrid Sparse Attention with CPU-GPU Parallelism for Long-Context Inference

2026-05-08 · Feiyu Yao, Zhixiong Niu, Xiaqing Li, Yongqiang Xiong 외 arxiv

Long-context inference increasingly operates over CPU-resident KV caches, either because decoding-time KV states exceed GPU memory capacity or because disaggregated prefill-decode systems place KV data in host memory. Al…